2.1.

Sample Preparation

Sepia eumelanin was purchased from Sigma-Aldrich (St. Louis, MO). For cuvette experiments, eumelanin was dissolved in 1-N $\mathrm{NaOH}$ aqueous solution and filtered through a $0.22\text{-}\mathrm{\mu }\mathrm{m}$ filter after $2\mathrm{h}$ of sonication. After that it was transferred to a quartz cuvette with 1-mm optical path length and diluted to 0.05 optical density (OD) at $775\mathrm{nm}$ (verified by a Cary 50 Bio spectrophotometer).

B16 mouse melanoma cells generally are highly pigmented, containing a fair amount of melanin. The cells used in the experiments were mounted on a slide with ${\mathrm{Vectashield}}^{\circledR }$ (Vector Laboratories, Burlingame, CA) after being cultured for 3 days.

A rhodamine 6G (Exciton, Dayton, Ohio) sample was prepared in methanol with a concentration of $40\mathrm{mM}$ . The solution was then transferred to a quartz cuvette with a 1-mm optical path length.

2.2.

Two-color Two-Photon Absorption/Excited State Absorption Measurements

The z-scan method is a traditional way of measuring nonlinear refraction and absorption of various materials. ^{22, 23} Since both nonlinear effects become obvious only when high pulse energy is used, usually an amplified laser system is required to perform z-scan measurements. The sensitivity is mainly limited by laser amplitude fluctuation noise, which is usually on the order of ${10}^{-3}$ . These high energy pulses cannot be used in biological systems because of photodamage concerns.^{24, 25} A more sensitive way to measure TPA has been demonstrated by using laser pulses with modest power directly from a mode-locked Ti: sapphire oscillator. The principle and experimental details for this method have been described elsewhere.²⁶ The basic idea is to use a high-stability pulse train with well-defined frequency components directly from a mode-locked laser. If a relatively slow sinusoidal amplitude modulation $(\mathrm{frequency}=f)$ is imposed on this pulse train, TPA creates sidebands $(\mathrm{frequency}=2f)$ in the transmitted light that would not be generated by linear processes. These sidebands permit very sensitive lock-in detection of the small TPA signals. This method can routinely measure ${10}^{-5}$ absorption changes with pulse energies of only a few picojoules. Further improvement is mainly constrained by the second harmonic of the $1f$ signal generated in the optoelectronic system, which interferes with the real TPA signal at $2f$ .

Instead of using a single modulated laser pulse train, TPA can also be measured by employing two synchronized pulse trains at different wavelengths; the optoelectronic nonlinearity constraint is then lifted and absorption changes on the order of ${10}^{-6}$ can be routinely measured. The apparatus of this two-color TPA measurement resembles the traditional pump-probe experimental setup. Pump-probe spectroscopy has been widely applied successfully to the investigation of ultrafast processes in condensed matter. The temporal evolution of different nonlinear processes (bleaching, stimulated emission, or excited-state absorption) can be studied by varying the time delay between the pump and probe pulses. Since our setup can detect very small transient absorption changes of the probe pulse with relatively low pulse energy, excited-state dynamics can be easily extracted as a nonlinear signature in microscopic studies.

The two-color two-photon measurement setup used in our experiments is shown in Fig. 1 . The laser system includes a mode-locked Ti: sapphire laser (Spectra Physics, Tsunami, $80\mathrm{MHz}$ , $100\mathrm{fs}$ ) and a synchronously pumped optical parametric oscillator (Spectra Physics, Opal, $80\mathrm{MHz}$ , $120\mathrm{fs}$ ). A small fraction of the Tsunami output was sent into a Mach-Zehnder interferometer to produce a pulse train with sinusoidal amplitude modulation at $2\mathrm{MHz}$ . This was achieved by directing beams of the two arms to two acousto-optical modulators (AOM) operating at $f_1=200\mathrm{MHz}$ and $f_2=202\mathrm{MHz}$ , respectively. The combined beam was therefore amplitude modulated at the difference frequency $f=f_1-f_2$ . The synchronous pulse train from the OPAL at $1300\mathrm{nm}$ was frequency doubled by a 1.5-mm-thick β-barium borate (BBO) crystal outside the cavity. Either the second harmonic at $650\mathrm{nm}$ or the fundamental residue at $1300\mathrm{nm}$ was used as the probe beam. The probe beam was combined with the pump beam at $775\mathrm{nm}$ on a dichroic mirror (760DCXR, Chroma, Rockingham, VT) and then sent to a home-built upright microscope ( $10\times \mathrm{objective}$ , $\mathrm{NA}=0.2$ , infinity corrected). Typical power used for our melanin solution measurements was $25\mathrm{mW}$ for $775\mathrm{nm}$ , $25\mathrm{mW}$ for $1300\mathrm{nm}$ , and $5\mathrm{mW}$ for $650\mathrm{nm}$ . The transmitted light was first collimated and then passed through another dichroic mirror to separate the pump beam and the probe beam, and the probe beam was then detected with a photodiode (Thorlabs, DET210 for $650\mathrm{nm}$ and DET410 for $1300\mathrm{nm}$ ) and a RF lock-in amplifier (Stanford Research Systems, SR844) at the modulation frequency. If any nonlinear process (two-color TPA, excited-state absorption, bleaching, or stimulated emission) occurred at the focus, the modulation of the 775-nm beam would transfer to the initially unmodulated 1300- or 650-nm beam, thus generating signals at $2\mathrm{MHz}$ on the lock-in amplifier.

Fig. 1

Schematic view of our two-color two-photon experimental setup. Interferometer is setup at $f1-f2=2\text{-}\mathrm{MHz}$ modulation. DM1, DM2: dichroic mirror. LPF: long pass filter. SPF: short pass filter. SHG: 1.5-mm-thick BBO crystal.

For cell imaging, a slightly revised setup was employed. We changed the interferometer to a single AOM modulating at $10\mathrm{MHz}$ on top of the 200-MHz carrier wave. Both the modulated 775-nm beam and the unmodulated 650-nm beam were traveling collinearly and sent to a home-built scanning microscope (scanning mirror from Cambridge Technology, $50\times \mathrm{NIR}$ objective, $\mathrm{NA}=0.55$ , infinity corrected). The power we used was $2\mathrm{mW}$ at $775\mathrm{nm}$ and $1\mathrm{mW}$ at $650\mathrm{nm}$ .

3.1.

Instrument Characterization with R6G Two-Photon Absorption

A R6G sample ( $40\mathrm{mM}$ in methanol in a 1-mm quartz cuvette) was first used to characterize our two-color two-photon setup. Since it has no linear absorption at the pump wavelength $\left(775\mathrm{nm}\right)$ , the only possible nonlinear process which could occur is two-color TPA by absorbing one pump photon and one probe photon simutaneously. The optical delay on the 775-nm beam path was scanned to record the two-color TPA signal. Figure 2 shows a typical TPA signal when the time delay between the 775 nm pulse and the 1300 nm pulse was varied. This signal is proportional to the cross correlation of two pulse intensities, since the TPA signal is instantaneous. Fitting it with a Gaussian function, we get a cross-correlation length of 238-fs FWHM. Assuming both pulses have an equal pulse duration, their pulse durations calculated from cross-correlation would be $168\mathrm{fs}$ . This means that our original pulses were slightly broadened after going through the AOM and the objective. The high modulation frequency we can use enables us to operate in the regime where the laser noise is negligible. The main noise source is from the detector and the lock-in amplifier. The noise level reading from our lock-in amplifier was less than $20\mathrm{nV}$ when the time constant was set to $100\mathrm{ms}$ . While the input signal from the photodiode monitoring the unmodulated beam is usually more than $20\mathrm{mV}$ , our system can easily detect transient absorption changes on the order of ${10}^{-6}$ .

Fig. 2

Two-photon absorption $(775+1300\mathrm{nm})$ signal from R6G.

3.2.

Two-Photon Absorption/Excited State Absorption Signal from Sepia Melanin

Melanins are heterogeneous polymers complexed with proteins. Though the basic monomer building blocks of these polymers are known, the detailed structures of the final products of melanin biosynthesis are still unknown.²⁷ Unlike other chromophores, melanins have structureless, monotonously deceasing absorption in the UV/VIS/NIR spectral region (Fig. 3 shows absorption spectrum of sepia melanin from 600 to $1300\mathrm{nm}$ ). TPF and TPA of melanins has previously been studied.^{17, 28} It is found that although TPF is rather weak (partly due to reabsorption), TPA is very large and provides good contrast in tissue imaging. While we can measure two-color TPA with higher sensitivity, the major advantage of our two-color two-photon setup is that time-resolved transient absorption behavior can be studied, similar to pump-probe spectroscopy. As a first illustration, Fig. 4 shows time-resolved transient absorption signal of sepia melanin solution when the pump is at $775\mathrm{nm}$ and the probe is at $1300\mathrm{nm}$ . Different processes can contribute to the transient absorption signal. TPA with a virtual intermediate state only happens when two pulses overlap in time; hence the decay signal we see out of the pulse overlap region is caused by excited-state absorption. Although stimulated emission or bleaching will give similar traces, they have exactly the opposite phase compared to ESA and TPA. Figure 5 illustrates the difference between these processes. The monotonously decreasing linear absorption of melanin suggests a quasi-continuous distribution of vibronic states above a certain level in the NIR range. The 775-nm pump beam will excite a certain fraction of melanin molecules to excited states. At $1300\mathrm{nm}$ , linear absorption from ground state melanins is negligible, but for those molecules excited by $775\mathrm{nm}$ to higher lying states, appreciable absorption will happen, which leads to decreased transmission, just as in TPA.

Fig. 3

Linear absorption spectra of sepia melanin.

Fig. 4

Two-color two-photon transient absorption (775-nm pump, 1300-nm probe) of sepia melanin.

Fig. 5

Schematic representation of different transient absorption processes. TPA and ESA lead to decreased transmission of probe, while bleaching and stimulated emission lead to increased transmission of probe.

The decay time of the ESA signal is a reflection of the excited-state lifetime after 775-nm excitation. The curve in Fig. 4 can be fitted with a double-exponential decay with time constants of $450\pm 50\mathrm{fs}$ and $3\pm 0.5\mathrm{ps}$ . The shorter component might have a contribution from TPA, which is instantaneous within the pulse overlap region. To differentiate between TPA and ESA contributions to the signals around zero delay, we studied the polarization dependence of these signals and the results are shown in Fig. 6 . It is known that the TPA cross section depends on the orientation of polarization of two exciting beams. ^{29, 30} If they make an angle $\theta$ , the two-photon cross section calculated with a simple geometric model is $\delta =A(2{\cos }^2\theta +1)$ .²⁹ The ratio of TPA cross section when $\theta =0\mathrm{deg}$ to that when $\theta =90\mathrm{deg}$ is 3. For complicated nonlinear molecules, the ratio becomes smaller. The R6G TPA signal has a ratio of 2.78, which is close to the theoretical value of 3. The melanin signal can be separated into two parts: a near-instantaneous part (within the pump probe pulse overlap) and a longer time decay part (out of the pump-probe pulse overlap). It is seen in Fig. 6 that there is almost no polarization dependence for the long time decay part, as expected, since it only has ESA contribution. Around zero delay, the signal ratio of $\theta =0\mathrm{deg}$ to $\theta =90\mathrm{deg}$ is 1.71, much smaller than R6G, which suggests that it has contributions from both ESA and TPA, with ESA playing a dominant role.

Fig. 6

Polarization dependence of TPA/ESA signal for R6G and pheomelanin samples. The angles shown are the polarization angle difference between the pump $\left(775\mathrm{nm}\right)$ beam and the probe $\left(1300\mathrm{nm}\right)$ beam.

Teuchner proposed that melanin TPF is caused by stepwise two-photon absorption,^{28, 31} but no further experiments were performed to provide direct proof of that process. Here by two-color two-photon absorption measurements, we can clearly see that an intermediate state exists and can facilitate the stepwise excitation process, which leads to efficient two-photon absorption and then fluorescence.

3.3.

Power Dependence of Melanin Two-Photon Absorption/Excited State Absorption Signal

To corroborate that our signal is from nonlinear absorption, the power scaling relationship with both pump and probe pulses was studied in detail. Figure 7 shows that the peak signal size (775-nm pump and 1300-nm probe) depends linearly on both pump and probe intensities for R6G and sepia melanin. The same relationship is found for the 775-nm pump 650-nm probe combination. Since the ESA signal depends linearly on the intensity of both beams, it has a nonlinear dependence on the total input intensity. This is a very important attribute, because it ensures that the ESA signal only comes from the focus where the total intensity is high. This nonlinear dependence permits us to do microscopy with inherent 3-D sectioning capability, just as does TPA and TPF. All the advantages in TPA/TPF imaging will be retained in the two-color two-photon ESA imaging.

Fig. 7

Power scaling relationship for both pump $\left(775\mathrm{nm}\right)$ and probe $\left(1300\mathrm{nm}\right)$ beams. When measuring power dependence on the pump beam, the probe power is fixed at $25\mathrm{mW}$ . When measuring power dependence on the probe beam, the pump power is fixed at $25\mathrm{mW}$ .

3.4.

Effects of Pulse Duration on Melanin Two-Photon Absorption/Excited State Absorption Signal

The difference in nonlinear behavior between TPA and ESA is reflected in their pulse duration dependence. When both excitation pulse durations are doubled, their peak powers will halve. Since the TPA signal is instantaneous and depends quadratically on the peak power but only linearly on the pulse duration,³² the resulting signal will also halve. However, this is not true for ESA. The excited-state lifetime ensures that any molecule that is excited by the pump beam will affect the probe beam independent of pulse durations as long as they are much shorter than the lifetime. The resulting signal only depends on the pulse energy of both pulses, regardless of their durations. Our experimental results are shown in Fig. 8 . The pulse durations were stretched by inserting a 1-cm highly dispersive ${\mathrm{TeO}}_2$ crystal into the beam path before the microscope. This increases the cross-correlation of the two pulses by a factor of 1.7, as can be calculated from the R6G TPA signal. It is seen from Fig. 8a that the peak amplitude of R6G TPA signal decreases 1.9 times, close to what we expected from previous analysis. However, the long time decay part of the ESA signals of sepia melanin, as shown in Fig. 8b, is almost unchanged. This means that the ESA signal is not affected much by broadening the pulse width. Changes around zero delay (with a ratio of 1.5) might be caused by the concurrent change of TPA signal or cross-phase modulation (XPM) signal (discussed in the next section) present in these samples. If ESA is used as a contrast in melanin imaging, we can use relatively long pulses $(<1\mathrm{ps})$ without reducing our signal size much. This has the advantage of reduced sample damage³³ and also fewer requirements on dispersion control in the system.

Fig. 8

Comparison of pulse duration effect in TPA and ESA signal. Measurements made with short duration and long duration pulses are represented by $▴$ and $∎$ marked lines, respectively.

3.5.

Interference from Cross-Phase Modulation

It was observed that a small interfering signal can also be generated with samples that do not have TPA, such as glass or water. The most likely origin of these interfering signals is XPM. XPM is an instantaneous third-order process, which results from the intensity dependent refractive index change (optical kerr effect). It is a well-known artifact in pump-probe spectroscopy,³⁴ but is mostly studied in the frequency domain. It can also happen in the spatial domain. When a high intensity pulse with a Gaussian spatial distribution enters a medium, it will modulate the refractive index of the medium with the same spatial pattern, which causes a lensing effect. When the second pulse passes the same region in the medium simultaneously, it will be focused or defocused by this instantaneous “lens,” depending on the sign of the optical kerr coefficient of the medium. By placing an aperture after the sample, the lensing effect from the nonlinearity of the medium will show up as a signal. Ideally, if we can collect all the transmitted light, we can only see absorptive effects but not dispersive effects like XPM. Due to the imperfection in collection optics, for example reflection from lenses, we can still see a small signal, even without an aperture when the focus is in the glass wall or water. Figure 9 clearly shows how differently the signals (775-nm pump and 650-nm probe) from R6G and water change with and without an aperture (The small bump for water signal might arise from XPM of glass wall). Apparently, in the case of water, an aperture will increase signal size significantly. However, for R6G, placing an aperture decreases the signal due to lower light collection efficiency. This suggests an extremely sensitive way of measuring optical nonlinearities. Generally speaking, the XPM interference from water would not affect our intepretation of the excited-state decay behavior of melanin, because the XPM signal is at least an order of magnitude smaller than the melanin ESA signal and disappears when two pulses are not overlapped.

Fig. 9

Comparison of aperture effect in TPA signal and XPM signal. Measurements made with and without aperture are represented by $▴$ and $∎$ marked line, respectively.

3.6.

B16 Cell Imaging with Melanin Excited State Absorption Contrast

As a demonstration of our ESA imaging capability, we imaged B16 mouse melanoma cells mounted on a glass slide with a home-built laser scanning microscope. Figure 10a shows the bright field image of a B16 cell. The scanned images of the same cell acquired at a series of interpulse delays are shown in Figs. 10b and 10c, with Fig. 10b being the $x$ channel image and Fig. 10c being the $y$ channel image constructed from the lock-in detection signal. The $x$ channel signal has the same phase as the pump beam, while the $y$ channel signal is in quadrature phase with the pump beam. The image sizes are $300\times 300\mathrm{\mu }\mathrm{m}$ or $512\times 512$ pixels. Image acquisition time is $52\mathrm{s}$ . The same color scale was used for all the scanned images. Our lateral resolution is better than $1\mathrm{\mu }\mathrm{m}$ . The signal-to-noise ratio (SNR) ranged from 30 to 50 for these images. The ESA signal decay can be clearly seen from the decrease of contrast in the $x$ channel as the interpulse delay increases (the round blue region in the zero delay image is most likely due to a XPM artifact from the glass slide, which disappears outside the pulse overlapping region). The maximum image intensity decreases more than five times by changing the pulse delay from 0 to $3.3\mathrm{ps}$ . However, the $y$ channel images remained almost the same, even at negative delay where no ESA signal is expected. It also shows distinctively different features from the $x$ channel images. We confirmed that this was not a linear absorption image, but was an image of some other nonlinear signature, which has a different phase relationship to the ESA signal. This suggests that the overall signal has contributions from different lifetime decays, which have different phase relations to the pump beam, similar to the frequency domain fluorescence.^{35, 36} The short lifetime decay $\left(3\mathrm{ps}\right)$ gives rise to a signal that has a zero phase ( $x$ channel signal), while the long lifetime decay gives rise to a signal that has a nonzero phase ( $y$ channel signal as well as part of an $x$ channel signal). The phase $\theta$ before zero delay (which should only have contribution from long lifetime decay) is around $76\mathrm{deg}$ . According to the formula of $\tan \left(\theta \right)=\omega \tau$ , with $\omega =10\mathrm{MHz}$ , the lifetime $\tau$ should be around $400\mathrm{ns}$ . However, this calculation is only a crude estimation, since the zero phase setting is not very accurate. While fluorescence lifetime can be well characterized with frequency-domain measurement, we do not have a direct comparison here, since fluorescence from melanin is very weak and is difficult to measure. In addition, the excited state measured with fluorescence might not be the same state as we are measuring with absorption. The origin of this extremely long-lived state is still under investigation. This new state might offer additional contrast in imaging melanin.

Fig. 10

(a): B16 mouse melanoma cell image under a bright field microscope; (b) and (c) $x$ and $y$ channel image of the same cell scanned with our two-color two-photon setup at various interpulse delays (indicated on the image, color online only).